Machine Learning and Artificial Intelligence (AI) are two of the buzzwords of the moment. In financial institutions, there is a lot of focus on emerging technologies, things like chatbots and intelligent processes.
But how do AI and Machine Learning apply to the ATM space, and how can they be used to make advancements?
Here at Renovite we are continuously looking at the data generated by ATMs and reflecting on what it could be used for – and how. One thing that AI is good at is determining ‘what is normal?’ and this is a key element for the operability of ATMs.
Some examples of models that have been built include:
- The ability to predict if the rate of cash being withdrawn equals what the cash management system predicted – i.e. will the ATM run out of cash? Many things can impact this. For example, if a nearby competing ATM fails, is that generating more traffic? Is the weather an influence on this ATM? Are the smaller denomination notes available on this ATM, if not, is this causing larger withdrawal amounts? These factors, together with predicting future values, is what machine learning excels at.
- Should we be seeing transactions from a particular ATM? It may be communicating fine, but is there a reason the ATM isn’t being used? My favourite example of when this happens is when various creatures i.e. snakes, rodents and other none-welcoming species move in and scare away customers. The lack of transactions when transactions are expected is a key part of identifying an occurrence like this and ensuring the efficiency of the monitoring process.
- Another use of AI is controlling the customer experience. Because AI algorithms will tend to store sequences of data and history, this can be used for recommendation engines at the ATM. We could use this in the ATM experience to target advertising messages – ensuring the adverts which get the best results are targeted towards customers. We are also reviewing how using AI can be used to create custom experiences for every customer based on the behaviour at ATMs.
- In Canada, the SickKids hospital is famous for using the data from health monitors together with AI techniques to predict when children are likely to have a life-threatening event, so staff can be alerted in advance. We use ATM logs to do the same thing for ATMs, capturing this data, and predicting when intervention may be required. For example, predicting card reader failure, electronic PIN pad or any other errors in the electromechanical parts in the cash dispensing process.
The examples above are just scratching the surface. Capturing and using data is just the first step, but in the age of cloud and on-demand processing power, harnessing and using the data is the first step in providing game-changing technology to continuously improve the profitability of ATM estates.
Interested in knowing more about our work in AI and Machine Learning? Get in touch.